Mounting a Heatsink

Bigman

Member
2007-11-19 4:26 pm
Hi guys,

I've just finished building my chip amp based on the LM4766. Sound is pretty good - I actually think my test speakers are letting the side down. I plan to get some decent speakers in the mix once I've built a box to move this thing.

One thing that is bothering me a bit though is how well I've mounted the heatsink.

During my testing, I can fairly easily hit the thermal overload protection. Whilst the chip is burning hot to the touch, the heatsink never reaches anything I would describe as 'hot' (though it is warm).

I can input roughly a 75% amplitude sine wave from my computer, which equates to approximately a 20V peak-to-peak output into 6ohm single-channel test speaker before it kicks in. I'm not sure how to work that out into power output, depending on how I work it out its somewhere between 8W and 33W per channel! If someone could shed some light on that, that would be great :)

I'm using a less-than-ideal set up for 6ohms (I designed with 8ohms in mind, and have a less-than-ideal power supply). My rails are +/-36.3V under load.

Anyway, I'm thinking that the heatsink should be getting a lot hotter than it is, and if it were dissipating the chips heat better that would improve the power headroom on my amp.

At present the heatsink is pressed up against the chip with a thermal pad between them and then glue-gunned together. I realise this solution is seriously ghetto - so fixing them together properly is my goal. One thing I am bothered to get right is having the heatsink electrically isolated from the chip, as the heatsink is going to external and so needs to be grounded like the rest of the case.

So any and all of your help, comments, suggestions etc are greatly appreciated and very welcome. Thanks in advance!
 

Dr_EM

Member
2006-10-18 12:00 pm
Swindon
It sounds like there may be a lack of pressure between the sink and the chip. To transfer the heat these must be pressed together very tightly so as to fill the tiny air gaps that otherwise occur. The surface of the sink must also be totally smooth and flat. Use a bolt through the mounting hole or a clamp across the chip to get the pressure needed. A "tap and die" set is needed to make a thread in the sink so that a bolt/machine screw can be driven into it.
 
Bigman said:
How can I ensure that the heatsink remains isolated if I were to screw/bolt through from the chip?

Use 'insulating bushes' for the screws and a non-conductive pad :)

See: http://www.rapidonline.com/producti...unting&tier4=Insulating+bushes&moduleno=31445 and http://www.rapidonline.com/products...er2=Semiconductor+Hardware&tier3=Thermal+Pads

[edit]
I found that the TO3-P sized pads work for the LM series chips, though the chip does overhang at the edges a tiny bit as the pad isn't quite wide enough. This wasn't a problem for me though. The two don't touch due to the pads thickness, and the heatsink surface is also anodized which I think provides some electrical isolation.
 
To test my Heatsink I made up an 8 ohm load by soldering two 50W resistors in series.


Then i used a shareware program to output a 1khz sine wave out my pc and had the oscilloscope connected to the input and the output.


I am using 35.44V power rails.

I then turned the volume up until my output reached the magic 28V mark (50W into 8ohm)

The wave started to get the top and bottom cut of at higher voltage so my max was 28V before the wave got distorted.

I left it at the full output for about a minute, the load resistors started to smell and got VERY VERY hot.

Heatsink got hot , but I could easily keep my hand on it , i didn't take any temps but heatsink was not even like a hot cup of coffee.

To mount chip onto alu heatsink i drilled on 3mm hole and used anough heat paste and a small selftapper screw to bind it down.

I even used the TF version of lm3886.

My heatsink is about (150 x 53 x 50) mm finned.

Tang
 
Hi Big,
if the chipamp (not the heatsink) gets too hot to touch when there is no input/output signal then you need better cooling.

If the chipamp temperature rises by more than 10Cdegrees above the quiescent temperature during power delivery in ordinary listening (whether quiet or party levels) then the heatsink is not dissipating sufficient heat.

If you do any serious power testing on continuous signals then the time period must be shortened so that the chipamp never overheats. You need to allow time for everything to cool down between each test run. I normally only do half power to full power tests for <=2seconds.

20Vac into 6ohms is 20*20/6=67W.
20Vpk (sinewave) into 6ohms is 20*20/2/6=33W.
Both of these test signals will raise the chipamp temperature very rapidly.

The chipamp must be securely clamped to the heatsink to maximise the heat transfer. Clamping is recognised as better than bolting for heat transfer.

Have a thorough read of the National datasheet. If you have any queries on content come back.
 

Bigman

Member
2007-11-19 4:26 pm
Thanks guys,

Just for now I've reattached it with a thermal pad and got it temporarily clamped. I will do a proper job of it once I've got the appropriate bits.

It still hits the overload protection around the 12Wrms mark playing a 100Hz sine wave, which I can probably attribute to high voltage rails and a 6ohm load. However, having had a good play with it with some music, I can seriously blast it (around 75Wrms peak through my single 6ohm speaker) without any protection kicking in. In fact, when giving it that much welly, it clips the output.

To my mind, that is great if it happy playing music at over-drive state without the overloads kicking in. Having run it for quite some time, the heatsink gets fairly warm, but not hot, and the chip feels roughly the same temperature - suggesting I now have a pretty reasonable thermal contact.

Next on the to-do list is to get two speakers connected in and do similar testing. After that I'll get the nice 8ohms on it :)

Thanks for the helps :)